By Cheskie Rosenzweig, MS, Aaron Moss, PhD, & Leib Litman, PhD
Most businesses can’t survive without conducting some research. What is our market share? Are our customers happy? Who is likely to buy this product? Questions like these are what lead businesses around the world to spend tens of billions of dollars per year on market research.
Regardless of whether you have a significant market research budget or one with very limited resources, it is of paramount importance for your business that your funds are spent efficiently and effectively. How do you do that? The first step might be recognizing when you do and do not need to gather your own data.
Not all market research requires a team of people to go out and gather data. Sometimes, your business has internal data, or you can use data other people have collected (known as secondary data) to answer your research questions. Internal data can help companies understand consumer behavior, and secondary data might help a company understand the market or its competitors.
But there are some questions no amount of internal or secondary data can answer. How do customers feel about our brand compared to others? How can we improve our product or service? Finding answers to questions like these requires talking to your customers or potential customers, and that means sampling people for the purpose of primary research.
As an example, imagine we lead the research team at a young company based in Minneapolis, Minnesota. Our company, aptly named SunVac, developed a new vacuum that runs on solar energy and never needs to be plugged in. As you might guess, we are excited that our hard work has come to fruition. We did it! We created an environmentally friendly vacuum with no more pesky wires to get tangled!
The problem we have now is that we aren’t sure how much our vacuum is worth on the open market. Although we have some secondary data on how much people will pay for wireless vacuums, we decide our product is sufficiently different from other models that we need to gather data to determine pricing sensitivity and the best way to market our product. The first step is determining who we need to sample.
Before embarking on any research project, it’s important to spend time clearly defining your objectives. Defining what you want to learn will guide your decisions about which source of data is best, how you should sample, and who you should sample.
Consider our company, SunVac. Our research team knows that we should conduct some studies investigating how much people will pay for our product and what kind of messages will convince people to buy it. From here, we need to define a target population for our studies, and while doing so, it is a good time to think about potential sources of sampling bias.
Is it important that our study represent certain demographic groups or people from various regions of the country? Should we make sure men and women are equally represented in the study? Does how much money people make influence whether they will buy our vacuum? Thinking about potential sources of bias can help us clarify who to sample.
Based on intuition and some secondary data, the research team at SunVac has a sense of who may have an interest in our product, who buy the product at different price points, and who respond to different marketing campaigns.
We decide we should sample people who may be in the market for a vacuum cleaner. We also decide it is important to collect data from people in various regions of the country to account for regional differences in environmental attitudes. If we limited our sampling to people in Minneapolis, we might end up with biased results, because Minneapolis is a city ranked cleanest in the U.S. and 6th-most eco-friendly in the world, meaning people in Minneapolis may value our product more than potential customers elsewhere. Finally, we consider data we have seen that married people vacuum more than single adults. We decide we should sample more married people than singles. So, our target sample is adults from various regions of the US who may be interested in buying a vacuum. Let us next consider where we could collect our sample.
Once you identify a target population, you need to form a plan to reach them and to gather your data. There are several related issues to consider.
Some people are harder to find as research participants than others. CEOs and managers are less plentiful than entry-level employees. There are fewer older adults online than younger adults. When forming a sampling plan, it is important to consider how hard it is to reach your target audience.
The amount of money budgeted for your project will affect your decisions about how to reach your target audience. For example, gathering a nationally representative sample based on probability sampling is often quite expensive. If it isn’t essential that your project be based on probability sampling, many researchers find it more affordable to collect a controlled sample that uses quotas to match to the U.S. census.
The amount of money you have budgeted for your project can also affect other considerations, such as where to find participants. Some online platforms allow researchers to do more of the work in data collection, which lowers overall costs. Other online platforms manage data collection for researchers, which adds to overall costs. How much money you have will influence the decisions you make.
How quickly you need your data will affect not only the total cost of your study, but also your decisions of how to sample. If you need the data quickly, then it doesn’t make sense to adopt a slow strategy like voluntary sampling or face-to-face interviewing.
When researchers need data quickly, they often turn to online sampling sources. The internet makes it possible to run faster and more affordable studies than many other methods of data collection.
The information you’re asking participants to provide may influence how and where you decide to gather data. Specifically, if you are looking for participants to engage in an hour-long task, during which they rate several products and provide detailed responses about each one, then you will probably get the best results from a crowdsourcing platform like Mechanical Turk. Crowdsourcing platforms allow you to control participant compensation, and by paying participants adequately for their time, it is possible to get data from crowdsourcing sites that participants from most online panels would never take the time to provide.
On the other hand, if you are gathering simple survey responses from participants, then there are many platforms that are suited to the type of data you seek to collect.
How might the questions above affect the research decisions we make at SunVac?
First, we know it’s relatively easy to reach our target audience. Any sizeable online panel should have access to adults from around the U.S. and allow us to target married couples.
Second, as a small company, we don’t have a massive budget for research. Because a random sample isn’t necessary for our research questions, we will gather a non-random sample and aim to control for potential sources of bias. For example, we will use quotas in our data collection to ensure we gather data from people of various ethnic and age groups.
Third, we want the data quickly. We know our competitors are close to developing a similar product, and we want to make sure our product hits the market first. As a result, we want to conduct our project within the next two weeks, meaning we should choose a sampling method and source that yield quick data.
Finally, our study asks participants to answer some questions about our product and to tell us which features of different marketing messages are most persuasive. Because our study isn’t too long or too demanding, we can consider a wide range of online panels with which to run our study.
To summarize, we know that most online panels will allow us to sample the people we are interested in, but we need our data quickly and we have a tight budget to stick to. The ideal platform for our project may be something like CloudResearch’s Prime Panels, or if we want to do some of the work ourselves, we might run the study on Mechanical Turk using CloudResearch’s MTurk Toolkit.
Now that we’ve built a sampling plan, we have to decide how many people to sample.
How many people you recruit into your study depends on your goals, the type of study you’re conducting, and how you plan to use your data.
If you’re conducting a survey, as our company, SunVac, is, then you need to consider a few factors when determining sample size. First, how large is the population you’re studying? As the size of the population you seek to understand grows, so does the number of people you need to sample. Our population for the SunVac project is quite large, encompassing nearly all adults in the U.S.
Second, how much inaccuracy are you willing to accept in the results? While your initial reaction may be “none,” it’s important to keep in mind that all sampling entails some margin of error. The question you have to answer is how important it is for your project to minimize the margin of error while balancing the increased costs of gathering a larger sample.
At SunVac, someone on our team has a background in statistical methods. She informs us it would be wise to run a conjoint analysis project asking people to rate the attractiveness of a series of descriptions of vacuum cleaners at different price points and with different features. She explains to us that it will take some time to design the survey itself, but she estimates that for appropriate statistical power to analyze the results among the different market segments we are interested in (region, relationship status, age groups), we will need data from 2,000 potential customers.
Now, you’re ready to find participants. The problem is that there is an overwhelming number of online options to choose from.
Depending on who you want to sample and what you want them to do within your study, online panels and crowdsourcing platforms both offer options for obtaining the sample you are interested in.
Online panels offer access to tens of millions of participants worldwide. When using online panels, researchers can easily target participants based on demographic characteristics, geographic location, psychographics and more. At SunVac, we could easily run our study using an online panel.
In addition to online panels, crowdsourcing platforms like Amazon’s Mechanical Turk are increasingly popular among market researchers. Crowdsourcing platforms give researchers more control over how their study is setup, how communication with participants takes place, and how much participants are compensated. Each of these features can be used to elicit more participant engagement than is typical in online panels.
If we decide at SunVac to conduct our study with an online panel, we will need the ability to collect high-quality data from a diverse sample of 2,000 adults, with a quota for a particular number of men and women who come from different age groups and regions of the country, and are either married or single. This means we will need a platform that allows us to selectively recruit 2,000 vacuum cleaner users for a 15—20 minute survey, and we want to make sure we collect good data from participants who are paying attention.
Ideally, what might happen next for SunVac, and hopefully to you, our reader, is that, in the process of researching how to find the best sample for your needs, you come to this website, read this page, and realize that CloudResearch has what you need. At CloudResearch, we have the ability to connect researchers with samples for nearly any project. In addition, we can provide advice for your data collection or gather the sample for you. Our solutions are tailored to your needs.
Why wait? Reach out today and see how we can help you achieve your research goals. Collect participants via Prime Panels or our MTurk Toolkit by signing up for a CloudResearch account, or ask for our assistance in designing your survey or sampling approach or for help with data collection or analysis today.